Machine Learning in Signal Understanding

Prof. Te-Won Lee

The University of California, San Diego

2003. 3. 26

Recently, there have been many approaches to solving complex signal processing problems using graphical models which provide a systematic and principled way to represent the problem and derive learning algorithms. Our research makes use of this framework in a variety of methods and applications. In signal understanding, we present generative models for speech signal analysis applied to speech enhancement and recognition. In building models for human vision, we derive methods for learning properties of early vision for grayscale and color image representation. Finally, for applications in medical informatics we present examples of generative data models in glaucoma diagnosis and prediction. The use of graphical models in signal representation provides a novel research direction allowing solving complex problems in a systematic manner.


This page is maintained by Ho-Jin Chung (
Last update: Mar. 26, 2003